Since the launch of estimated in-store visits in December, 2014, the vast majority of savvy Internet marketers have been hesitant to trust and utilize the data in an effort to optimizer their campaigns – and for good reason. To date, Google has treated the release of estimated total conversions (ETC) as somewhat of a black box, simply stating that in-store visits were based primarily upon user proximity to an advertiser’s location on Google Maps (when location history is activated on their phones). With the release of products such as Blueprint that are furthering the capabilities of DSPs to tie digital advertising to in-store visits and sales, additional pressure has been placed on Google to shed light on the accuracy of ETC data – in addition to its partnerships with the likes of Acxiom and Datalogix. Intuitively, as investment in mobile advertising continues to grow it will become ever more important that Google can decipher between actual in-store visits and users simply being in proximity to a store location.
Speaking @HeroConference in Portland this year, Surojit Chatterje – Google’s PM for Mobile Search – shared more explicit details concerning how exactly AdWords estimates in-store visits (and how accurate these estimates really are). As we’ve assumed, there are a variety of factors that go into in-store metrics apart from simple location history:
- Google Maps Street View/Google Earth data
- Wi-Fi signal strength of particular retail locations
- GPS signals
- Query data
- Mapping of coordinates/borders of stores – supposedly of millions, globally
- Visit behavior of users
To Chatterje’s point, the strength/validity of ETC data is backed by the sheer breadth of user information at Google’s disposal – information that is continually sampled to inform and improve data models that can understand signals and predict outcomes. The type of machine learning that is used to support estimated total conversions is backed by a panel of one million users who opt-in to provide on ground location history to Google on a daily basis – this level of data informs Google if an in-store visit indeed occurred. Of particular interest, Google’s model does not automatically consider a user entering a store a “visit;” alternatively, search patterns and queries are analyzed to determine if a visit was most likely informed/influenced by a Search ad.
Ultimately, in-store transaction data is meant to be partnered with store transaction data – allowing large advertisers to build 360 degree attribution models, down to the keyword level. This deep level of attribution is able to speak volumes concerning campaign planning and bidding strategies; however, automating this level of data into AdWords is currently not a capability (primarily, only very large advertisers/retailers are able to take advantage of Google’s partnership with Datalogix).
The update to estimated total conversions falls in line with #StepInsideAdWords news concerning Search Funnels/Attribution – primarily the ability to select an attribution model per conversion type and integrate that data into reporting and your automated bidding strategies (available later this year, including the ability to integrate cross-device conversions into bids). The update to Attribution integration in AdWords is paramount in regards to advertisers who do not have access to Google Analytics Premium and/or Adometry (both of which have this capability) – the end goal being actionable attribution data available to all advertisers, regardless of spend or account size.